Optimizing wild bee surveys for conservation management: A comparison of survey methods

Statewide wild bee surveys are essential baselines for monitoring population trends and informing conservation. Despite a recent accumulation of statewide surveys, we lack consensus on sampling methods, limiting data comparability across states. We examined effects of sampling frequency, geographic replication, and land protection status on bee richness, functional representation, and community composition for a state wide pan trapping survey. We generated 46,533 specimens representing 272 species across 143 sites sampled 2–4 times monthly. Simulations that iterated 3, 2, or 1 randomly selected samples per month per site estimated significantly lower species richness, likely reflecting a high number of singletons. Specialists and cleptoparasites, which were overrepresented among singletons, were disproportionately affected by reduced sampling. Species richness of parks and preserves was similar to that of residential properties, but the species composition differed with ~30 species more likely to be found in parks than residential properties. We contrast the pan trap results with a subsequent netting survey that targeted the host plants of specialist bees. Netting surveys occurred in different years and sites, but over the same months, geographic ranges and habitats. We conducted 1,200 15-min netting surveys that yielded 4,299 specimens of 196 species. Though netting documented fewer unique species than pan trapping (46 versus 97), discovery efficiency (measured by # species x # specimen^-1) was three times higher and documented more specialist species, as well as cavity nesters and cleptoparasites. Netting more efficiently captured conservation-relevant components of bee communities than pan trapping.